o wi3@sddlmZmZddlZddlmZddlmZmZmZm Z m Z ddl m Z ddl mZmZmZdd lmZeeZGd d d e ZGd d d eZGdddejZGddde ZGdddeZGdddeZgdZdS))OptionalUnionN)nn)LlavaCausalLMOutputWithPastLlavaForConditionalGeneration LlavaModelLlavaModelOutputWithPastLlavaPreTrainedModel)ACT2FN)auto_docstringis_torchdynamo_compilinglogging)VipLlavaConfigc@ eZdZdS)VipLlavaModelOutputWithPastN__name__ __module__ __qualname__rrj/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/transformers/models/vipllava/modular_vipllava.pyr%rc@r)VipLlavaCausalLMOutputWithPastNrrrrrr)rrcs*eZdZdeffdd ZddZZS)VipLlavaMultiModalProjectorconfigcstt|jtr dnt|j}tj||jj |j d|_ tj ||jj |j j dd|_t|j|_tj |j j |j j dd|_dS)Nr)epsT)bias)super__init__ isinstancevision_feature_layersintlenr LayerNorm vision_config hidden_sizeprojector_layernorm_epsprojector_layernormLinear text_configlinear_1r projector_hidden_actactlinear_2)selfrnum_feature_layers __class__rrr .s   z$VipLlavaMultiModalProjector.__init__cCs,||}||}||}||}|SN)r)r,r.r/)r0 hidden_statesrrrforward=s    z#VipLlavaMultiModalProjector.forward)rrrrr r6 __classcell__rrr2rr-src@r)VipLlavaPreTrainedModelNrrrrrr8Err8c@seZdZ ddejdeeeeeffddZ e            ddej dejdeej deej d eeejd eejdeeeeefd ee d ee d ee dee deej deeeffddZdS) VipLlavaModelN pixel_valuesr"csv|dur|n|jj}|j|ddt|tr$j|ddddf}nfdd|D}tj|dd}||}|S) aW Obtains image last hidden states from the vision tower and apply multimodal projection. Args: pixel_values (`torch.FloatTensor]` of shape `(batch_size, channels, height, width)`) The tensors corresponding to the input images. vision_feature_layers (`Union[int, list[int]]`): The vision feature layer, or the list of indexes of the layers to select the vision feature. Returns: image_features (`torch.Tensor`): Image feature tensor of shape `(num_images, image_length, embed_dim)`). NT)output_hidden_statesrcs&g|]}j|ddddfqS)Nr)r5).0index image_outputsrr ds&z4VipLlavaModel.get_image_features..)dim) rr" vision_towerr!r#r5torchcatmulti_modal_projector)r0r:r"image_featuresrr>rget_image_featuresJs  z VipLlavaModel.get_image_features input_idsattention_mask position_idspast_key_values inputs_embeds use_cacheoutput_attentionsr; return_dictcache_positionreturnc  Ks| dur| n|jj} | dur| n|jj} | dur| n|jj} |dur$|n|jj}|du|duAr4td|dur@|dur@td|durJ||}|dur|j||d}||jjk d}| | |j }t s|||kr||jjk}|jd|jd}td|d || |j |j}|||}|jd |||||| | d | d | }t|j|j|j|j|dur|ndd }| r|S|S)z vision_feature_layers (`Union[int, list[int]]`, *optional*): The vision feature layer, or the list of indexes of the layers to select the vision feature. Nz:You must specify exactly one of input_ids or inputs_embedszdYou cannot specify both pixel_values and inputs_embeds at the same time, and must specify either oner:r"rArrz6Image features and image tokens do not match: tokens: z , features T) rJrKrLrMrNrOr;rPrQ)last_hidden_staterLr5 attentionsimage_hidden_statesr)rrOr;use_return_dictr" ValueErrorget_input_embeddingsrHimage_token_id unsqueeze expand_astodevicer numelsumshapedtypemasked_scatterlanguage_modelrrTrLr5rUto_tuple)r0rIr:rJrKrLrMr"rNrOr;rPrQ lm_kwargsrGspecial_image_maskn_image_tokensn_image_featuresoutputsoutputrrrr6isb    zVipLlavaModel.forwardr4) NNNNNNNNNNNN)rrrrD FloatTensorrrr#listrHr LongTensorTensorbooltuplerr6rrrrr9Is^        r9c!@seZdZ ddejdeeeeeffddZ              ddej dejdeej d eej d eeejd eejdeeeeefd eej d ee dee dee dee deej deeej fdee effddZdS) VipLlavaForConditionalGenerationNr:r"cCs|jj||dS)NrS)modelrH)r0r:r"rrrrHsz3VipLlavaForConditionalGeneration.get_image_featuresrrIrJrKrLrMlabelsrNrOr;rPrQlogits_to_keeprRcKs| dur| n|jj} | dur| n|jj} | dur| n|jj} |dur$|n|jj}|jd||||||| || | d| d |}|d}t|trLt| dn|}| |dd|ddf}d}|durm|j |||jj j d}t |||j|j|j|jdS)a vision_feature_layers (`Union[int, list[int]]`, *optional*): The vision feature layer, or the list of indexes of the layers to select the vision feature. labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. Example: ```python >>> import torch >>> from PIL import Image >>> import requests >>> from transformers import AutoProcessor, VipLlavaForConditionalGeneration >>> model = VipLlavaForConditionalGeneration.from_pretrained("llava-hf/vip-llava-7b-hf", device_map="auto", torch_dtype=torch.float16) >>> processor = AutoProcessor.from_pretrained("llava-hf/vip-llava-7b-hf") >>> prompt = "A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.###Human: \n{}###Assistant:" >>> question = "Can you please describe this image?" >>> prompt = prompt.format(question) >>> url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/compel-neg.png" >>> image = Image.open(requests.get(url, stream=True).raw) >>> inputs = processor(text=text, images=image, return_tensors="pt").to(0, torch.float16) >>> # Generate >>> generate_ids = model.generate(**inputs, max_new_tokens=20) >>> processor.decode(generate_ids[0][len(inputs["input_ids"][0]):], skip_special_tokens=True) The image features a brown and white cat sitting on a green surface, with a red ball in its ```NT) rIr:rJrKrLrMrNr"rOr;rPrQr)logitsrt vocab_size)lossrvrLr5rUrVr)rrOr;rWr"rsr!r#slicelm_head loss_functionr+rwrrLr5rUrV)r0rIr:rJrKrLrMr"rtrNrOr;rPrQrurfrjr5 slice_indicesrvrxrrrr6sH4 z(VipLlavaForConditionalGeneration.forwardr4)NNNNNNNNNNNNNr)rrrrDrlrrr#rmrHrnrorprqrr6rrrrrrsh         rr)r9rrr8)typingrrrDr(transformers.models.llava.modeling_llavarrrrr activationsr utilsr r rconfiguration_vipllavar get_loggerrloggerrrModulerr8r9rr__all__rrrrs    q f